A Letter from the Founder

Hey everyone 👋, I'll get right to the chase — this letter isn't generated by AI, it's actually written by me, a human, and gramma corrected by AI. I want to share on a personal note what this product is trying to solve and what problems I've faced throughout years in the software engineering space, now watching AI reshape our industry.

We all know the drill. Sprint plannings, retros, user stories, bug tickets. A lot of effort goes into planning a sprint, and even though AI can help with many things, it still requires business decisions on what to build and what goes next. The process has always been requirements from the business team, transcribed into tickets by analysts or project managers, then translated into engineering items we track on our boards.

The issue I've faced many times is that project managers can't truly make tickets and give estimates independently — lacking technical context makes it harder. They need an engineer to assess the load, and technical expertise is usually required for a well defined ticket with the right timelines. When minor bug fix tickets stack up, timelines get pushed for work that in reality won't take more than a few hours, but results in days lost in sprints and engineers doing trivial work.

Synlets tries to address this. I've made a tool where you can have an agent hooked up to your codebase, have a conversation about upcoming work to help PM's generate a well defined ticket with full codebase context, and estimate what needs to be done and how. Plus once you make that ticket you can assign it to an AI agent directly and let it do the work.

Another issue I've faced personally and seen many engineers face — we get basic tickets all the time. Fix a typo, add this button, add a small feature, add security groups to infrastructure, add this user to terraform. What do we do now? We paste it into Cursor, Claude Code, or Codex, it modifies the code, we check it within seconds and go "yep, that looks about right" and make a pull request. I've worked as a principal engineer and I could see that I could spend time delivering more value rather than doing those basic things. After chatting with many project managers I came to conclusion that they would also prefer to get the easy stuff done immediately, without having to disturb sprint.

There will always be a gate before a PR goes into production — a human has to review it and verify it works. I would personally rather review a PR from a feature branch / ephemeral environment and verify that tests pass, it works, and merge it, rather than doing the prompting myself, checking locally, pushing, evaluating tests, re-iterating on failures, and only then merging. It's better for AI to do the work, push it with automated tests, and get engineers notified: this ticket has been implemented, tests passed, please review. If all good, we just approve it.

For highly complex features we will still need people to think creatively, and when products are truly complex it requires human comprehension to steer AI in the right direction. Building Synlets with AI was made easier, but I can tell you with certainty if it's not steered right it gets messy fast. But for simple and mid level tickets that don't require deep context, AI can do just fine. Engineering work should focus on delivering high value items, and PM's should have their own agents to handle minor/mid level items without disturbing sprint workflow.